p.1275

Information Impact and Allocation Rules in Auctions with Affiliated Private Values: A Laboratory Study

John H. Kagel
Ronald M. Harstad
Dan Levin

Abstract

In affiliated private value auctions, each bidder has perfect information regarding his/her own value for the object at auction, but higher values of the item for one bidder make higher values for other bidders more likely. We report on a series of experiments examining three key implications of these auctions: (i) in a first-price auction, public information about rivals' values increases expected revenue, (ii) an English auction increases expected revenue compared to a first-price auction, and (iii) a second-price auction is isomorphic to an English auction. In examining these issues, we compare predictions of some ad hoc bidding models with Nash equilibrium predictions. In the first-price auction experiments, Nash equilibrium bidding theory organizes the data better than either of two ad hoc bidding models. Public information about others' valuations does increase average revenue, but the increase in revenue is smaller and less reliable than predicted under risk neutral Nash equilibrium bidding. Lower average revenue might be attributed to risk aversion, while the high variability is attributed to a sizable frequency of individual bidding errors relative to the theory. Bidding theory precisely organizes English auction outcomes after a brief initial learning period. The dominant strategy equilibrium does not organize second-price auctions nearly as well, as market prices persistently exceed predicted prices. The difference between English and second-price outcomes is attributed to effects of different information flows, inherent in the structure of the two institutions, on eliminating bidding errors. Revenue impacts of these two institutions, relative to a first-price auction, are examined in light of observed bidding patterns.